How Karaca Turned an AI Shopping Assistant Into a Business Transformation Engine

When Karaca CEO Fatih Karaca decided to expand his company’s use of artificial intelligence beyond digital marketing optimization, he was not simply chasing a technology trend. He was making a deliberate, structured bet on operational transformation at a company that had spent over five decades building its identity on quality craftsmanship and white-glove customer service. The results, documented in a McKinsey case study, are striking: a 5x higher conversion rate through the AI-powered shopping assistant AIDA, a 2x improvement in advertising return on investment, and the identification of more than ten actionable AI domains from a pool of nearly 200 candidates.

These numbers tell a compelling story. But for business leaders reading beyond the headline metrics, the more important narrative is about how Karaca got there—and what that process reveals about the real requirements of enterprise AI adoption.

The McKinsey case study positions Karaca’s journey as a model of AI-enabled operational transformation in the retail sector. This is a fair characterization, but it also invites critical scrutiny. The Karaca story is genuinely instructive not because it confirms that AI delivers results but because it illustrates the specific conditions under which it does. Those conditions have less to do with algorithm selection and more to do with leadership alignment, strategic discipline, and an unglamorous commitment to data quality. Understanding what Karaca did well, what questions the case study leaves unanswered, and what the broader research literature suggests about AI transformation in retail gives business leaders a far richer framework for action than the headline metrics alone.

AI Transformation Requires Strategic Prioritization (Not Blanket Adoption)

The central thesis of the Karaca case is deceptively simple: by partnering with McKinsey and QuantumBlack and systematically working through nearly 200 potential AI application domains, the company identified the interventions most likely to deliver meaningful improvements in savings, efficiency, value creation, and customer experience. This is not a story about AI doing magic. It is a story about disciplined prioritization.

That distinction matters enormously. A recurring failure mode in enterprise AI adoption is the tendency for organizations to pursue AI applications opportunistically, chasing whatever use case appears most exciting or most visible at a given moment. Karaca’s approach was deliberately the opposite. The company began with extensive cross-functional workshops spanning category management, marketing, operations, sales, legal, finance, and HR. The goal was not to find places where AI could be applied but to find places where it should be applied, based on potential impact.

This sequencing—strategy before technology—is something that management researchers have identified as a critical differentiator between organizations that realize sustained value from AI and those that generate isolated pilots without enterprise-wide impact. A 2023 MIT Sloan Management Review study found that companies with clearly defined AI strategies tied to specific business outcomes were significantly more likely to report measurable financial returns from their AI investments than those pursuing a more experimental or ad hoc approach. Karaca’s methodology appears to reflect this discipline.

The choice to begin with AIDA, the consumer-facing shopping assistant, is itself worth examining. Fatih Karaca has described it as a product with the potential to redefine the shopping experience, comparable to the impact that mobile apps had on consumer behavior more than a decade ago. This framing is ambitious, but the underlying logic is sound. Consumer-facing AI applications offer a rare combination of attributes: they are measurable through hard conversion metrics, they generate rapid feedback loops, and they create organizational momentum by producing results that internal skeptics can see and respond to. Karaca and McKinsey appear to have recognized that a visible, customer-oriented success would build the internal confidence needed to pursue the broader AI roadmap.

The Conversion Rate Claim That Demands Context

The 5x higher conversion rate attributed to AIDA is the most immediately striking figure in the case study. It is also the one that most requires contextual interpretation. The comparison baseline—unaided e-commerce sessions on the existing mobile app—is an important qualifier. Assisted sessions, in which a user actively engages with a recommendation tool, have historically produced higher conversion rates than unassisted browsing across a wide range of e-commerce contexts. The question for leaders is not whether the AIDA-assisted sessions converted better, but whether the magnitude of improvement—and its durability over time—justifies the investment and the architecture of the solution.

On the cost side, the case study reports that Karaca was able to reduce the cost per chatbot session by 97.5% prior to launch. This figure, if accurate, is remarkable and speaks to a sophisticated optimization effort. Generative AI inference costs have been declining across the industry as model providers reduce pricing and as organizations learn to engineer more efficient prompts and retrieval architectures. The fact that Karaca’s teams were actively working to reduce session costs before going to market suggests a level of operational rigor that is often absent in first-generation AI deployments.

A critical observation worth raising—one that the case study does not fully address—concerns the long-term retention and satisfaction metrics associated with AI-assisted shopping. Conversion rate is an important short-term signal, but it does not capture whether customers who purchase through an AI-assisted experience return at higher rates, report greater satisfaction, or develop stronger brand loyalty. Research from Salesforce’s State of the Connected Customer report (2023) found that 73% of consumers expect companies to understand their unique needs and expectations, but 56% say they generally feel treated like a number rather than a person. An AI assistant that converts effectively but fails to create genuine personalization depth may produce a short-term lift without the long-term loyalty that justifies the investment.

Data Quality: The Unsung Hero of AI Transformation

Perhaps the most practically useful insight in the Karaca case study comes not from the headline metrics but from a remark by board member Emre Karaca: “Data quality is the core of everything. Because while AI is smart, if data is not correct, available, or reachable, AI responses will be half complete.”

This observation cuts directly to a challenge that continues to derail AI initiatives at organizations across sectors. A 2022 IBM Global AI Adoption Index found that 35% of companies identified data complexity and limited AI expertise as the primary barriers to AI adoption. Many organizations discover too late that their data infrastructure—accumulated over years of legacy systems, siloed operations, and inconsistent data governance practices—is fundamentally inadequate for the demands that AI systems place on it.

Karaca’s response was to establish formal protocols for improving data quality prior to each AI project launch. This is a mature organizational posture, one that treats data readiness as a precondition for AI deployment rather than as an afterthought. For business leaders considering their own AI roadmaps, this is arguably the most transferable lesson from the Karaca experience. Before asking what AI can do for your organization, ask what your data can do for your AI.

Leadership and Change Management as Strategic Assets

The Karaca case study contains a detail that is easy to overlook but critically important for any leader managing an AI transformation: the role of McKinsey senior partner Can Kendi’s weekly breakfast meetings with CEO Fatih Karaca in troubleshooting challenges and generating new ideas. This informal touchpoint—alongside structured A/B testing and fact-based analysis—served as a mechanism for sustained leadership engagement at the highest level.

This matters because AI transformations are not primarily technology problems. They are organizational change problems in which technology happens to play a central role. Emre Karaca’s observation that teams became excited about AI once they saw the results of rigorous testing is a textbook illustration of what organizational behavior researchers call evidence-based change management. Rather than asking skeptical employees to trust in the technology on faith, Karaca’s leadership created the conditions for teams to form their own evidence-based conclusions. This approach is more durable than top-down mandates because it builds genuine internal capability and conviction rather than mere compliance.

This dynamic is consistent with findings from a landmark study by Erik Brynjolfsson and colleagues at MIT, which found that firms that combined AI investment with complementary investments in worker skills and organizational redesign generated substantially greater productivity gains than those that deployed AI without addressing the human and organizational dimensions of the transformation. Karaca’s cross-functional workshops, its leadership-driven change management approach, and its emphasis on team capability building all suggest an awareness of this principle, even if the case study does not explicitly frame it in those terms.

The Case for AI-Driven Marketing Optimization

The reported 2x improvement in advertising ROI through AI-driven, cross-platform digital marketing optimization is the second major headline result in the case study. Orphoz leader Berk Topcu’s characterization of the impact as “like having a thousand people analyzing your platform performance, 24/7” captures something real about the nature of AI-enabled marketing optimization.

Modern digital advertising ecosystems—spanning search, social, display, video, and increasingly connected TV and retail media networks—generate data volumes and decision frequencies that no human team can optimize manually. AI systems can continuously adjust bid strategies, creative rotation, audience targeting, and budget allocation across platforms based on real-time performance signals. The result is not just efficiency but a form of continuous optimization that was practically impossible before machine learning-driven bidding and attribution systems became widely available.

The doubling of advertising ROI is plausible in this context, particularly if Karaca had previously been relying on more manual or rules-based optimization approaches. Research from Google’s own studies of AI-powered Performance Max campaigns has shown average conversion value increases of 18% over standard Shopping campaigns, while independent analyses of programmatic advertising adoption have documented ROI improvements in the 20% to 40% range. A 2x improvement is at the high end of what the published literature would suggest is typical, but it is not implausible for an organization moving from a relatively unsophisticated baseline to AI-driven optimization.

What the Case Study Leaves Unanswered

For all of its instructive value, the Karaca case study has limitations that leaders should recognize. Published by McKinsey, which served as Karaca’s implementation partner, the case study is by nature a success narrative. It does not discuss projects that failed to deliver expected results, AI domains that were deprioritized after initial exploration, or the organizational costs and disruptions associated with the transformation. The 97.5% reduction in chatbot session costs, for instance, raises questions about what the original cost baseline was and what the absolute cost figures look like at scale across millions of users.

The case study also does not address the ethical and governance dimensions of deploying consumer-facing AI at scale, including how Karaca manages customer data privacy, how it ensures that AIDA’s recommendations are fair and free from bias, and what recourse customers have when the AI assistant gives incorrect or misleading guidance. These are not peripheral concerns. As generative AI applications proliferate in retail, regulatory scrutiny of AI-driven consumer interactions is intensifying across multiple jurisdictions, and organizations that fail to build robust AI governance frameworks alongside their technical capabilities are accumulating legal and reputational risk.

Key Takeaways for Leaders

For business leaders outside the homeware retail sector, the Karaca story offers several transferable principles:

• Strategic prioritization matters more than speed of adoption. The discipline of working through 200 potential domains to identify the ten most impactful is an organizational capability, not a one-time exercise. Building the internal capacity to rigorously evaluate AI opportunities, assess data readiness, and estimate realistic impact is a competitive advantage in its own right.
• A consumer-facing application is often the right place to start. Not because it is easiest, but because it creates the feedback loops and organizational momentum that sustain broader transformation. Karaca’s decision to lead with AIDA proved both strategically sound and commercially productive.
• Data quality is a strategic imperative. It is not a technical problem to be solved by the IT department alone. It requires executive ownership and cross-functional protocols. Organizations that treat data governance as a prerequisite for AI deployment rather than a consequence of it will move faster and capture more value.
• Change management is not optional. The evidence base for AI adoption increasingly shows that the organizations capturing the greatest value from AI are those that have invested simultaneously in technology, talent, and organizational redesign. The tools alone are not sufficient.
• Success metrics should extend beyond what’s immediately measurable. Conversion rates and advertising ROI are essential signals, but durable advantage from AI comes from the organizational capabilities, customer relationships, and data assets that accumulate over time. Karaca’s CEO has said that the biggest value from AI may still be ahead. That forward-looking posture—treating early results as a beginning rather than a destination—is perhaps the most important leadership lesson the case study has to offer.

Karaca’s journey from a family glassware workshop in 1973 Istanbul to an AI-enabled, nearly billion-dollar retailer operating across 43 countries is a story about many things: generational leadership, brand building, and geographic expansion. The AI chapter is still being written. But the approach Karaca and its leadership team have taken—patient, structured, rigorous, and deeply attentive to both the technical and the human dimensions of transformation—provides a useful model for any organization serious about moving from AI experimentation to AI-driven enterprise value creation.